Monday 8 June 2009

Measure for measure

Sky have a revealing little graphic for their World T20 coverage. As they preview the next three batsmen in, they show a single stat - their strike rate.

Murdoch companies make their fortunes by taking complex pieces of information and making them simple; it's worked across tabloid newspapers and television channels for forty years. They have a nose for what the punter likes, and a saucy strike rate is the sexiest new stat in cricket. It's easily understandable, and it's a comparatively big number. 

T20 has produced an obvious demand for new ways of measuring player data. Batting average doesn't cut it for T20, it's too skewed. In Test cricket a benchmark of, say, 40 works across the top six. It's universal enough to apply to each of those batting positions. But in T20, the opportunity for numbers one, two and three batsmen to build higher averages than those coming in after over number 10 is far greater. Equally, a number six or seven in a decent side may get a very high proportion of not outs. Either way, the average loses meaning, and stats are all about meaning.

So strike rate measures the speed of scoring, which is meaningful. But it feels as though it needs a supplement, a stat that tells you over what period that speed has been sustained. Average would be the obvious measure, but again it doesn't feel accurate enough. The skewing mentioned above is one reason. Another is that high team scores in T20 generally come from using a variety of batsmen to optimum effect: blazing hitting is usually suited to shorter individual innings, and individual wickets hold less value than they do in longer games.

So what measure should sit alongside strike rate? Maybe it should be average duration of innings, or average number of balls faced. Something like: 's/r 150.33, av b/f 28.1'

Now like KP, maths is not my strongpoint. I guess that multiplying those two stats out would result in something very much like an average. But batting effectiveness might be more clearly stated via the method above. 

Over to you, stattos.


Q said...

"a saucy strike rate is the sexiest new stat in cricket"

As far as T20 is concerned that is.

The same thing is being shown on prime sports, I think the coverage is the same world over..

Another thing I've noticed is when they show the incoming batsman's record, instead of 50s and 100s, they show 30+..

So 30+ is the new benchmark for T20.. should batsmen raise their bats when they get to 30?

I think a SR and Avg balls faced per innings is a great measure..

Probably will come in as we go along..

As the dot balls measure has come in for the bowlers.. when the bowling figures are shown, there is no column for maidens, its over, runs, wickets, dot balls, which I think is apt for T20.

The Old Batsman said...

Yes, I think 30 is probably fair, but it doesn't seem impressive!

Meant to mention the dot balls in the post. I think that is a good stat - although it can make you look bad as well as good - Brett Lee had 5 in his spell of 1-56, so he essentially conceded 56 in 3 overs...

Q said...

I think thats the idea with dot balls.. Lee going for 56 of 19 deliveries.. fastest 50 conceded? Haha..

30 doesn't look impressive now.. give T20 another 5-10 years and it may start looking impressive..

An Econ rate of 6.5 doesn't look good at all.. but in T20 its actually very impressive..

The eye and the mind will have to get used to these new sets of benchmarks for T20..

The Old Batsman said...

The eye and mind - spot on, Q. Actually, another batting measure might be dot balls, too - ie how many you don't score off...?

Q said...

I think that is already there.. when they show the break up of a batsman's score they usually show:

Dilshan 37*

0s: 2, 1s: 3, 2s: 2 4s: 6 6s: 1

But they don't do it more often..

achettup said...

Nope, avg balls faced - if I assume it to be the number of balls faced per dismissal - multiplied by strike gives you the average times 100. Again if it is per innings, it ends up being skewed since not everyone will get to face the same number of balls per innings.

I like strike rate * batting average, I think it tends to even things out, though as with the previous case batsmen higher up the order get better chances, but then that has always been the case in limited overs matches.

What they might end up doing is classifying different kinds of batsmen and using different weights based on the position. Batsmen 1-3 might have higher emphasis on average while batsmen 5-7 on strike-rate. 4 is Michael Clarke and nobody should care about him.

David Barry said...

I sort of agree with Achettup. Average times strike rate is probably the best simple measure of a batsman in limited-overs cricket, and it's what I'd use if I needed a stat to compare batsmen and didn't have several weeks of research time trying to work out something better.

Having said that, I'm usually happy just seeing the strike rates. Especially in T20I's, where they haven't played many games, the average is much less reliable than the strike rate.

(I mean that in a sense of prediction - if you see a guy with a strike rate of 150, you know he scores fast. If you see a guy with an average of 20, you don't really know if he's a good batsman who's been unlucky with his slogging in the ten T20I's that he's played, a genuine 20-average player, or a lower-order slogger who had a lucky day and smacked 50 off 20 once.)

So really I look at the strike rate, and to guess how long he'll stay in the middle, I use the rough idea in my head of how good the batsman is. That idea is more informed by 50-over cricket than 20-over internationals.

Russ said...

I'm inclined to go back to the Duckworth/Lewis model, because although it is complex, it is probably the most mathematically advanced understanding of the nature of a batting resource, and the changing balance between balls faced and a batsman's wicket.

Runs per D/L resource used also has the nice property that it can be used to compare bowlers as well.